基于分形理论和数学形态学的木材表面缺陷识别的图像处理

在木材表面缺陷识别过程中,能否精确提取缺陷轮廓是提高分选准确率的重要因素。采用分形理论和数学形态学进行板材缺陷图像分割和边缘提取,能有效的克服传统方法中缺陷图像提取受背景纹理影响大的不足,为后续的木材表面缺陷的识别打下坚实的基础,提高木材的使用率,对木材表面缺陷识别具有实际意义。     

基于改进GAC模型的木材表面缺陷图像分割

为提高对木材表面缺陷图像分割的准确率,对木材表面缺陷图像采用传统GAC模型算法和改进GAC模型算法进行多组对比试验,与此同时研究改进算法中迭代步长、迭代次数、常数速度、反差参数等参数对木材表面缺陷图像分割结果的影响。通过试验验证了改进GAC模型算法的可行性、快速性和准确性,能够克服传统的GAC模型欠分割的缺点。     

基于OTSU算法与数学形态学的木材缺陷图像分割

在木材分选过程中,图像缺陷分割技术占有重要的地位,能否精确提取缺陷轮廓会直接影响到分选的准确率.本文讨论提取木材表面缺陷图像的方法,应用OTSU算法与数学形态学相结合的方法对缺陷图像进行分割,最终提取出缺陷边缘.实验表明,经过OTSU算法和数学形态学进行图像分割,最后得到的木材缺陷图像更加清晰、连贯,提高了图像的可视性和准确性.     

基于邻接自适应谱聚类的木材表面缺陷分割算法

针对人工分割木材表面缺陷的烦琐性和阈值分割算法对缺陷像素信息衡量的不稳定性,提出了一种基于邻接自适应谱聚类的木材表面缺陷分割算法。算法以简单线性迭代超像素(simple linear iterative cluster, SLIC)为基础,对缺陷图像进行预处理,融合木材缺陷的纹理特性和超像素块间的距离尺度,并采用邻接自适应谱聚类进行分割;缺陷分割初步完成后,通过变异系数衡量缺陷块中像素信息的离散程度进行再次分割,克服初次分割结果的过分割问题;考虑木材表面缺陷形态学上的封闭性,将2次分割图像进行合并,继而用邻接扫描法对次分割图形进行填充,最终对木材表面缺陷进行分割界定。考虑木材表面缺陷种类的多样性,选取了虫眼、死节、活节等缺陷图像进行分割对比试验,相较于OTSU阈值分割算法,本研究算法在单个和多个木材表面缺陷分割方面,类别平均像素准确度(mean pixel accuaracy, MPA)分别提升4.69%,14.23%,平均交并比(mean intersection over union, mIoU)分别提升33.27%,33.43%。本研究算法能够更加准确地将木材表面缺陷从复杂背景中...     

3种不同模型对木材表面缺陷图像分割算法的比较

木材表面缺陷会严重影响木材的质量和使用价值,因此对木材表面缺陷图像分割的研究有利于提高木材的利用率。本文分别对红皮云杉含有虫眼、活节、死节3种典型木材缺陷的图像采用改进的C-V模型、改进的GVF Snake模型和改进的GAC模型进行分割试验,对3种改进算法的复杂程度、分割时间、分割结果的完整性以及抗噪性进行对比和分析。结果表明,改进的GAC模型算法较为优越,其分割算法简单,运行时间短,缺陷分割效果较好,抗噪性强。而改进的C-V模型算法、改进的GVF Snake模型算法的分割效果和抗噪性最差,不宜作为3种木材表面缺陷图像的分割算法。     

基于LS-SVM的木材表面缺陷网格化检测

针对图像分割的复杂性和局限性,作者提出一种基于最小二乘支持向量机(LS-SVM)的木材表面缺陷网格化检测方法。首先将木材表面图像划分成互不重叠的矩形块,然后依次计算每个矩形块图像的特征向量,用于描述各个矩形块图像,其特征向量由颜色特征和纹理特征等参数共同组成。最后将特征向量归一化后送入LS-SVM分类器,利用特征向量的相似度来进行缺陷的定位和识别。实验结果表明,该方法可有效进行木材表面缺陷检测,检测准确率超过93%。     

基于图割算法的木材表面缺陷图像分割

木材表面缺陷分割的研究能够有效提高木材的利用率,节约现有木材资源,缓解森林资源短缺的压力。为了更好地对板材表面的节子和虫眼进行快速有效地分割,论述了基于图割算法的图像分割方法(Graph Cuts)及其改进方法(Grab Cuts)的原理。针对传统Graph Cuts算法只能针对灰度图像进行分割、运行时参数的选择比较复杂,并且存在该算法效率和精度较低的缺陷,采用这两种方法分别对3种木材表面缺陷活节、虫眼和死节图像进行分割实验。为了验证Grab Cuts方法的适用性,用含有多个缺陷目标的木质板材图像做了图像分割验证。结果表明:缺陷图像的目标和背景的种子点选取直接影响Graph Cuts算法的分割结果,Graph Cuts算法的计算效率较低,分割时间较长,对相邻像素间的区分度较差,分割结果不理想。改进后的Grab Cut算法是迭代的Graph Cuts,该方法虽然在图像分割前也需要人工画定初始化矩形框,但操作相对简单,分割结果能够得到完整的闭合缺陷区域边界,且不受木材表面缺陷的类型、数量、尺寸和缺陷形状的影响,分割效果好,分割速度快,抗噪性强,对灰度图像和彩色图像都可使用。     

数学形态学在木材表面缺陷图像分割后处理中的应用

介绍了数学形态学的基本思想和运算。针对木材表面缺陷图像分割效果不完善的问题,提出基于数学形态学的图像后处理方法,包括应用数学形态学的填充操作、形态滤波以及形态梯度边缘检测等。经实验验证,应用数学形态学进行图像后处理,增强了木材缺陷图像分割结果的可视性和准确性。     

木质板材表面缺陷自动检测技术研究

木质板材表面缺陷检测系统采用高亮光源提供照明,线阵CCD相机实时获取其表面图像,利用工控机在线处理数据并根据板材缺陷的检测分析选择相应的生产工艺和加工方法,以提高木材的出材率和生产的自动化程度。该系统可完整提取板材表面缺陷,利用改进的差影法对图像进行分割,以便对图像特征进行提取。实验表明,该数据处理方法能够准确地提取板材表面缺陷的信息。     

木材缺陷图像的灰度形态滤波

介绍数学形态学中的灰度形态滤波方法,给出对木材缺陷图像进行闭、开运算处理后的仿真图像实例,利用木材缺陷图像截面灰度分布对处理效果进行分析.结果表明:采用灰度形态滤波方法对木材缺陷图像进行滤波处理,可有效去除木材缺陷图像中的噪声,改善图像的视觉效果,提高后期缺陷边缘检测的精确度,从而证明该方法对木材缺陷图像进行滤波处理是可行的.     

55mm厚红桉板材气干特性的研究

对55mm厚红桉板材气干特性与规律的研究结果表明,当红桉板材的含水率在30%以上时,在广州的2、3、4月份进行气干比较合适,在其它月份则应采取适当保护措施,以防止产生干燥缺陷;板材气干到含水率25%时即进行窑干,具有较好的经济效益;板材窑干前应进行调湿处理,以降低含水率梯度,从而有效避免表裂的发生.     

Method of check analysis of microwave-modified wood

A photo-based method for statistical image analysis of microwave (MW)-modified timber was developed and applied to test sets of Radiata pine and eucalypt hardwoods (Messmate and Mountain Ash). The method is based on filling the checks in timber with stain solution and analysing the wood surface. Bethel impregnation process (in a pressurised treatment vessel) was used to produce uniform surface staining of the test samples. Image processing was automated and the number of manual operations (the human decision-making process) was minimised. A computer program that automates thresholding and a program that repairs the threshold image were written. The software allowed larger images to be analysed and reduced image processing time. The described method produced reliable check measurements and statistics for softwoods. Though suitable for estimating the quality of individual MW-modified hardwood specimens, the method was found to be inadequate for statistical analysis of eucalypt hardwoods because of collateral staining of vessels accompanied with a high level of stain penetration into the wood tissues.     

基于数字图像处理的樟子松锯材分级研究

为了解决传统锯材检测方法效率低、劳动强度高和人为因素影响大的问题,笔者运用数字图像处理的方法对樟子松锯材等级进行检测,实现樟子松锯材等级检测的标准化和程序化。根据樟子松锯材缺陷与天然木纹颜色存在显著差距的特点,利用Otsu's方法的最佳全局阈值处理锯材图像得到最佳分割阈值,获得锯材的二值图像,得到缺陷位置像素坐标,将缺陷坐标与锯材实际尺寸对应,计算出相应锯材的净划材尺寸和出材率,完成对锯材等级的划分。利用该方法对75片试件进行表面质量检测和等级划分的准确率为94.67%,分级程序的运行时间约为1.793 s。该方法在保证检测准确率的情况下不仅可以减少检测时间、降低人工劳动强度、提高检测准确率,而且解决了人工检测主观性强的问题。     

基于Faster R-CNN的实木板材缺陷检测识别系统

我国木材资源有限,为了提高木材的利用率,采用机器视觉来实现木材缺陷快速而稳定的检测,不仅可以克服人工检测的低效率和木材缺陷识别的低准确率,而且对提高木材加工企业的智能化水平具有重要意义。为了高效、快速、准确地进行无损检测,采用深度学习方法,建立了一种基于快速深度神经网络的实木板材缺陷识别模型。首先采用Resnet V2结构对采集到的实木板材缺陷图像进行特征提取,然后应用该模型对节子、孔洞等实木板材缺陷进行训练学习,最后构建了Faster R-CNN检测框架,并使用tensorflow开发平台对节子、孔洞等实木板材缺陷进行预测输出。具体选取了2 000块杉木样本,通过旋转对原始的实木板材图像进行数据扩充,扩充后图像的80%作为训练集,20%作为验证集来进行仿真。仿真结果表明,该模型对实木板材节子缺陷检测正确率为98%,对实木板材孔洞缺陷检测正确率为95%,验证了将深度学习算法应用于实木板材缺陷检测中的有效性。     

Effect of cross-sectional dimensions on bow and surface checking of sugi (Cryptomeria japonica) boxed-heart square timber dried by conventional kiln drying

The effect of cross-sectional dimensions on bow and surface checking were investigated, using the boxed-heart square timber of two sugi cultivars with dimensions 80, 120, and 140 mm and length 1.9 m taken at two different heights above the ground. The smaller cross-sectional timber tended to have larger bow, less surface checking, and larger dimensional shrinkage. However, the drying defects were different between the cultivars and sampling heights, depending on the shrinkage properties of the juvenile wood and the heartwood proportion in the core part of the stem. The bow was larger in the smaller cross-sectional timber in which the longitudinal shrinkage was large in the juvenile wood. Surface checking was more prominent in larger cross-sectional timber containing sapwood in its outer part, which suggested the surface checking was induced by drying stress, owing to large moisture gradients between the heartwood and sapwood. The cross-dimensional shrinkage of the timber was larger in timber with larger tangential shrinkage.     

基于声波的木材孔洞位置检测的探讨

针对不同位置孔洞缺陷的木材,进行室内敲击试验。试验收集木材受敲击时发出的声音信号,并对信号进行数据分析,研究具有不同缺陷的木材在不同敲击条件下共振频率的变化,并探讨此共振频率与木材缺陷的对应关系。结果表明,木材在不同缺陷条件下（无孔洞／有中间孔洞／有端部孔洞）的共振频率有明显不同,且这种对应关系仅与缺陷情况有关,与敲击力度和敲击方位无关。通过检测木材的声音共振频率来判定木材孔洞位置的方法是可行的。     

装饰木材美感度的评判

以65种具代表性的装饰木材为评价对象,采用大众评判法获得公众对每种装饰木材的喜好度值,用多元数量化理论Ⅰ模型建立喜好度与14个装饰木材表面要素间的回归关系,并用影响装饰木材美感度的颜色、色差、表面硬度、表面粗糙度4个主要因子建立了装饰木材美感度模型.结果表明,装饰木材的美感度随着表面硬度的增大而增大,随着表面粗糙度和色差的增大而减小,深暗色装饰木材具有较高的美感度.     

Wettability of Sitka spruce timber stored in a Scottish sawmill yard

Kiln-dried and air-dried Sitka spruce battens were exposed to outdoor weathering within stickered, close packed and wrapped packs and at the upper surface of these packs. A four-week exposure period was used in James Jones and Sons Ltd. sawmill yard near Aboyne, NE Scotland. The wettability of batten samples removed from the packs of timber was determined by a drop contact angle method using a video tape technique. Timber exposed to air drying before weathering in the experimental packs generally had a greater wettability than kiln-dried timber at the end of the exposure period. However, kiln-dried timber stored in stickered packs developed a greater wettability than similarly stored air-dried timber, suggesting that the kiln drying process increased the susceptibility of timber to developing increased wettability. Close-packed timber protected from sunlight and rain water, by covering the upper surface of the pack with microporous plastic sheeting, maintained a lower wettability than unprotected close-packed or stickered timber. Fungal spoilage was unlikely to develop on timber stored in pack types which developed high timber wettability, the storage conditions leading to an increase in wettability were adverse to fungal development. Kiln-dried timber stored in stickered packs in the sawmill yard may become particularly vulnerable to fungal attack if subsequently exposed to damp conditions in service. Received 15 October 1998     

Soft measurement of wood defects based on LDA feature fusion and compressed sensor images

We proposed a detection method for wood defects based on linear discriminant analysis(LDA) and the use of compressed sensor images. Wood surface images were captured, using a camera Oscar F810C IRF camera,and then the image segmentation was performed, and the defect features were extracted from wood board images. To reduce the processing time, LDA algorithm was used to integrate these features and reduce their dimensions. Features after fusion were used to construct a data dictionary and a compressed sensor was designed to recognize the wood defects types. Of the three major defect types, 50 images live knots, dead knots, and cracks were used to test the effects of this method. The average time for feature fusion and classification was 0.446 ms with the classification accuracy of 94%.